Skip to content
AX Principles

Principles

Agent Experience Principles

What your product owes the machines that now use it.

25 of 25 principles

AX-SYS-01

Make your purpose machine-readable

Context and comprehensionFoundational
AX-SYS-02

Serve a markdown version of key content

Context and comprehensionIntermediate
AX-SYS-03

Keep machine docs in sync with the product

Context and comprehensionIntermediate
AX-SYS-04

Let agents authenticate without human-only gates

Access and identityFoundational
AX-SYS-05

Give agents a distinct, least-privilege identity

Access and identityIntermediate
AX-SYS-06

Return actionable authorization errors

Access and identityFoundational
AX-SYS-07

Publish a capability index agents can find

Discovery and navigationIntermediate
AX-SYS-08

Tell agents the next action, not just a menu

Discovery and navigationIntermediate
AX-SYS-09

Expose a machine-readable map of your resources

Discovery and navigationIntermediate
AX-SYS-10

Expose only the tools the agent needs

Tools and capabilitiesIntermediate
AX-SYS-11

Write tool descriptions for the model, not the human

Tools and capabilitiesFoundational
AX-SYS-12

Treat tool descriptions as a trust boundary

Tools and capabilitiesAdvanced
AX-SYS-13

Make consequential actions explicit and reversible

Tools and capabilitiesIntermediate
AX-SYS-14

Design for graceful failure

Recovery and reliabilityFoundational
AX-SYS-15

Be idempotent and predictable where it matters

Recovery and reliabilityIntermediate
AX-SYS-16

Persist state across multi-step workflows

Memory and stateAdvanced
AX-SYS-17

Tell agents when things change

Notifications and changeIntermediate
AX-SYS-18

Be triggerable and context-passable by other agents

Orchestration and handoffAdvanced
AX-SYS-19

Define roles and handoff contracts for multi-agent flows

Orchestration and handoffAdvanced
AX-SYS-20

Make agent activity visible

Observability and governanceAdvanced
AX-SYS-21

Bound autonomy with scopes, limits, and spend caps

Observability and governanceIntermediate
AX-SYS-22

Never trust agent-generated markup in sensitive contexts

Observability and governanceAdvanced
AX-SYS-42

Enforce agent access parity with other mediums

Access and identityFoundational
AX-SYS-43

Design agent-first onboarding — let agents start before humans create accounts

Access and identityIntermediate
AX-SYS-44

Expose feedback endpoints so agents can report on context quality and task outcomes

Observability and governanceIntermediate